Validating Belief States of an AI System by Sentiment Analysis and Controversy Detection

1. A method comprising: providing, by a computer, a question answering service; detecting, by the computer, a negative sentiment of a user to an answer transmitted by the computer to a device associated with the user; performing, responsive to detecting the negative sentiment and by the computer, controversy detection using a plurality of data sources and based upon a topic related to the answer; determining, by the computer and based upon the controversy detection, that a controversy related to the topic exists; adding, by the computer, a new belief state to the question answering service based on the controversy; and transmitting, by the computer, an updated answer to the device, wherein the updated answer is based on the new belief state.

2. The method of claim 1 , further comprising: receiving, by the computer, a question from the user device; determining, by the computer, a first belief state of the question answering service, wherein the first belief state is determined based on the question received from the device; and transmitting, by the computer, a first answer to the device based on the first belief state, wherein the first answer and the answer for which the negative sentiment is detected are the same.

3. The method of claim 2 , wherein the first belief state and the new belief state are semantically contrary to one another.

4. The method of claim 1 , further comprising: receiving, by the computer, a question from the user device; determining, by the computer, a first belief state of the question answering service, wherein the first belief state is determined based on the question received from the device; and transmitting, by the computer, a first answer to the device based on the first belief state; detecting, by the computer, a first negative sentiment of the user to the first answer; responsive to detecting the first negative sentiment, selecting, by the computer, a second belief state of the question answering service based on the question received from the device; and transmitting, by the computer, a second answer to the device based on the second belief state, wherein the second answer and the answer for which the negative sentiment is detected are the same.

5. The method of claim 4 , wherein the first belief state and the second belief state are alternative belief states of the question answering for the received question.

6. The method of claim 4 , wherein the second belief state is utilized by reinforcement learning algorithms of the question answering service.

7. The method of claim 1 , wherein detecting the negative sentiment comprises: receiving, by the computer, from the user device a response to the answer transmitted by the computer to the device associated with the user; and analyzing, by the computer, the response to the negative sentiment.

8. The method of claim 1 , further comprising determining, by the computer, when content exists that is semantically contrary to the answer.

9. The method of claim 1 , wherein the adding the new belief state increases a number of belief states in the question answering service as a potential answer to a question to which the answer is a response.

10. The method of claim 1 , wherein the new belief state is utilized by reinforcement learning algorithms of the question answering service.

11. A system, comprising: a processor programmed to initiate executable operations comprising: providing a question answering service; detecting a negative sentiment of a user to an answer transmitted by the system to a device associated with the user; performing, responsive to detecting the negative sentiment, controversy detection using a plurality of data sources and based upon a topic related to the answer; determining, by the computer and based upon the controversy detection, that a controversy related to the topic exists; adding a new belief state to the question answering service based on the controversy; and transmitting an updated answer to the device, wherein the updated answer is based on the new belief state.

12. The system of claim 11 , wherein the processor is programmed to initiate executable operations further comprising: receiving a question from the user device; determining a first belief state of the question answering service, wherein the first belief state is determined based on the question received from the device; and transmitting a first answer to the device based on the first belief state, wherein the first answer and the answer for which the negative sentiment is detected are the same.

13. The system of claim 12 , wherein the first belief state and the new belief state are semantically contrary to one another.

14. The system of claim 11 , wherein the processor is programmed to initiate executable operations further comprising: receiving a question from the user device; determining a first belief state of the question answering service, wherein the first belief state is determined based on the question received from the device; and transmitting a first answer to the device based on the first belief state; detecting a first negative sentiment of the user to the first answer; responsive to detecting the first negative sentiment, selecting a second belief state of the question answering service based on the question received from the device; and transmitting a second answer to the device based on the second belief state, wherein the second answer and the answer for which the negative sentiment is detected are the same.

15. The system of claim 11 , wherein the first belief state and the second belief state are alternative belief states of the question answering for the received question.

16. The system of claim 11 , wherein the processor is programmed, in order to detect the negative sentiment, to initiate executable operations further comprising: receiving from the user device a response to the answer transmitted by the computer to the device associated with the user; and analyzing the response to the negative sentiment.

17. The system of claim 11 , wherein the processor is programmed to initiate executable operations further comprising: determining when content exists that is semantically contrary to the answer.

18. The system of claim 11 , wherein the adding the new belief state increases a number of belief states in the question answering service as a potential answer to a question to which the answer is a response.

19. The system of claim 11 , wherein the new belief state is utilized by reinforcement learning algorithms of the question answering service.

20. A computer program product, comprising: a computer readable storage medium having program code stored thereon, the program code executable by a data processing system to initiate operations including: providing, by the data processing system, a question answering service; detecting, by the data processing system, a negative sentiment of a user to an answer transmitted by the computer to a device associated with the user; performing, responsive to detecting the negative sentiment and by the data processing system, controversy detection using a plurality of data sources and based upon a topic related to the answer; determining, by the data processing system and based upon the controversy detection, that a controversy related to the topic exists; adding, by the data processing system, a new belief state to the question answering service based on the controversy; and transmitting, by the data processing system, an updated answer to the device, wherein the updated answer is based on the new belief state.